deletions | additions       diff --git a/Method.tex b/Method.tex  index 0836ada..12802e6 100644  --- a/Method.tex  +++ b/Method.tex 
...
The frames containing single photon events were processed with the ThunderSTORM \cite{Ovesny2014} superresolution imaging plug-in for ImageJ, whose settings are explored in this paper. Due to computational memory restraints, the USAF test pattern data was processed in six 5,000 image stacks. The data is first processed to detect the photon events from the noise background, before implementing a localization algorithm to detect the centre of the photon event. Subsequently, a sub-pixel localization algorithm is used on the data to fragment the pixels into 13$\times$13 grids to allow the centre of photon events to be located with greater resolution.  For all the data, the parameters used for the camera settings include a pixel size of 80.0~nm and 36 photoelectrons per A/D count. The base level varied between image stacks due to fluctuations in the apparatus temperature, and was set to the average minimum grey value for the image stack which ranged between 100 A/D counts and 140 A/D counts. For all the data, a wavelet (b-spline) image filter was applied with a b-spline order of 3 and a b-spline scale of 2.0. For the approximate localization algorithms, centroid of connected components (CoCC) was used with a peak intensity threshold (PIT) of 2*std(Wave.F1) and the watershed algorithm enabled for the USAF test pattern data, and a PIT of 1.5*std(Wave.F1) with the watershed algorithm enabled for the cells data. Local maximum (LM) was applied using a PIT of 2.5*std(Wave.F1) with a 4-neighbourhood connectivity and non-maximum suppression (NMS) was used with a PIT of 2*std(Wave.F1) and 1 pixel dilation radius {{in radius. ***in  the end only CoCC was used for the data in this paper, is it necessary to mention LM and NMS?}}. NMS?***  For the sub-pixel localization algorithms, integrated Gaussian (IG) PSF was set to a 3 pixel fitting radius, a 1.6 pixel standard deviation and the least squares (LS) fitting method and radial symmetry was applied with a 2 pixel estimation radius. To achieve good photon detection with adequate FPN in a very short period of time, the data can be processed with the Fast Approximation (FA) settings described below. To achieve optimal photon detection, the resolution of overlapping events and a lower FPN, the Maximum Resolution (MR) settings can be applied to the data, however this produces a substantial increase in processing time. For the FA processing approximate localization algorithms, Gaussian PSF was set to a 2 pixel fitting radius with a 1.0 pixel standard deviation and the Maximum Likelihood (ML) fitting method, and for the MR processing, Gaussian PSF was set to a 7 pixel fitting radius with a 1.0 pixel standard deviation with the Maximum Likelihood (ML) fitting method.   When the multiple-emitter fitting analysis (MFA) was applied in conjunction to the PSF gaussian or IG algorithms, it was used with a maximum of 1 molecule per fitting region and a model selection threshold (p-value) of 10^{-6} for the FA processing, or a maximum of 2 molecules per fitting region and a model selection threshold (p-value) of 10^{-6} for the MR processing. For USAF test pattern data processed using the MR settings with MFA enabled, thunderSTORM's remove duplicates post-processing tool was applied with a distance threshold of 160~nm and an 'intensity>4000 filter' "intensity>4000" filter  was used. For cells data, the remove duplicates tool was also applied with a distance threshold of 160~nm but with an 'intensity>3000' "intensity>3000"  filter. No post-processing was required for data processed using FA settings.